Enable chat completition/embeddings Openai Endpoint in Guidance Server

[doomgrave]

The Guidance server was a brilliant addiction to serve guidance, but at the moment the model hosted can work only via guidance.
Would be possible to enable the standard llama-cpp server on "http://localhost:8392 /v1" ad example so it can be also used with Langchain/Openai so we can load a model once and use it in a completely flexible way!